News Intel Tiger Lake SuperFin-ishes AMD Ryzen 4000 In Latest Benchmarks

InvalidError

Titan
Moderator
Good to see competition, but could someone tell Intel to simplify its model numbers. ;)
The new-style model numbers for 10nm parts are simple enough: two numbers for generation, two numbers for models within that generation, 'G' to indicate that the chip has graphics and then numbers to indicate what IGP it has.

Decoupling the CPU part of model numbers from the graphics part clears the way for scaling them independently later with 2.5D/FOVEROS.
 

jimmysmitty

Champion
Moderator
In ST with a clock advantage and possibly faster memory I can see the CPU outperforming Zen. As for some of the tasks I can only guess that it might be DL Boost, the AI learning they have on CL chips, or possibly something similar to QuickSync that uses the iGPU when performing these tasks to speed up performance. Otherwise I can't see, if the application is able to use multiple threads, 4 cores beating 8. And I highly doubt TL has 2x the performance per core of ICL.
 

Avro Arrow

Distinguished
In ST with a clock advantage and possibly faster memory I can see the CPU outperforming Zen. As for some of the tasks I can only guess that it might be DL Boost, the AI learning they have on CL chips, or possibly something similar to QuickSync that uses the iGPU when performing these tasks to speed up performance. Otherwise I can't see, if the application is able to use multiple threads, 4 cores beating 8. And I highly doubt TL has 2x the performance per core of ICL.
I agree 100%. Intel CPUs winning in Intel-conducted tests are not the least bit meaningful.
 

escksu

Respectable
Aug 8, 2019
450
157
1,860
0
In ST with a clock advantage and possibly faster memory I can see the CPU outperforming Zen. As for some of the tasks I can only guess that it might be DL Boost, the AI learning they have on CL chips, or possibly something similar to QuickSync that uses the iGPU when performing these tasks to speed up performance. Otherwise I can't see, if the application is able to use multiple threads, 4 cores beating 8. And I highly doubt TL has 2x the performance per core of ICL.
Its entirely possible for 4 cores to beat 8. This is due to thermal and power throttling. Although Ryzen has 8 cores, its still limited by its 25W tdp.
 

jimmysmitty

Champion
Moderator
Its entirely possible for 4 cores to beat 8. This is due to thermal and power throttling. Although Ryzen has 8 cores, its still limited by its 25W tdp.

Yes but the TL chip is also limited to 28W so the odds of just 3W allowing that much performance advantage is slim. I wouldn;t be surprised if they don't have something that can use the iGPU. QuickSync took everyone by surprise.
 

InvalidError

Titan
Moderator
Yes but the TL chip is also limited to 28W so the odds of just 3W allowing that much performance advantage is slim.
Well, between TL having 50% higher base+boost clocks, ~20% higher IPC and most software not scaling anywhere near linearly to eight cores, it does not seem that far-fetched that a fast quad could be signfiicantly faster than a much lower-clocked octo in a fair number of cases.
 
Reactions: JarredWaltonGPU

jimmysmitty

Champion
Moderator
Well, between TL having 50% higher base+boost clocks, ~20% higher IPC and most software not scaling anywhere near linearly to eight cores, it does not seem that far-fetched that a fast quad could be signfiicantly faster than a much lower-clocked octo in a fair number of cases.
Well, between TL having 50% higher base+boost clocks, ~20% higher IPC and most software not scaling anywhere near linearly to eight cores, it does not seem that far-fetched that a fast quad could be signfiicantly faster than a much lower-clocked octo in a fair number of cases.
Yes but still it would feel a bit closer to me. In a highly multithreaded application that can scale well.

I still would not be surprised if Intel plans to utilize the iGPU if possible. It only makes sense.
 

Dsplover

Prominent
Feb 23, 2020
31
11
545
1
In ST with a clock advantage and possibly faster memory I can see the CPU outperforming Zen. As for some of the tasks I can only guess that it might be DL Boost, the AI learning they have on CL chips, or possibly something similar to QuickSync that uses the iGPU when performing these tasks to speed up performance. Otherwise I can't see, if the application is able to use multiple threads, 4 cores beating 8. And I highly doubt TL has 2x the performance per core of ICL.
Agree as my apps are just now stepping up so my 3700X can can be used for core parking and multicore optimizations.

Love AMD and a Intel when they fight for my coin.
Give me a fast low latency quad for desktop/HEDT, call it whatever you want as long as the cache is big and fast.
 

InvalidError

Titan
Moderator
Yes but still it would feel a bit closer to me. In a highly multithreaded application that can scale well.
Another thing that could further skew results in TL's favor is use of DDR5: less power going to the CPU's memory controller leaves another 1-2W extra package power for cores and IGP. Increased memory bandwidth would also help with IGP performance and possibly boost IPC on the CPU side too thanks to reduced memory controller contention between IGP and CPU.
 

Liquidrider

Reputable
Nov 25, 2016
3
0
4,510
0
Hold the phone I saw Fatality, and closing out with this
"It's too early to proclaim Intel as the undisputed winner, though. As we've learned in the past, manufacturer benchmarks aren't exactly the most accurate. "

I feel I just got Intel benchmarked :eek:.

Kidding of course great article.
 

dddjks

Distinguished
Feb 6, 2010
2
0
18,510
0
Leaked promotional videos reveal performance comparisons between Intel Tiger Lake and AMD Ryzen 4000-series processors.
Intel Tiger Lake SuperFin-ishes AMD Ryzen 4000 In Latest Benchmarks : Read more
I think this is premature article. Consider what PASSMARK has to say about the Intel vs Ryzen comparison.
Link: CPU Compare - PASSMARK cpubenchmark.net
(BTW this link is dynamic, you may see more samples in the near future.)

Something is wrong, even for a single benchmark, a CPU Mark of 7979 for Intel vs 17579 for Ryzen is a HUGE difference between "competitive" CPUs.
 
Last edited:

escksu

Respectable
Aug 8, 2019
450
157
1,860
0
Yes but the TL chip is also limited to 28W so the odds of just 3W allowing that much performance advantage is slim. I wouldn;t be surprised if they don't have something that can use the iGPU. QuickSync took everyone by surprise.
OK, I am not sure if this can be considered "cheating". Its because these low power Intel CPUs have a boost period where they can use 45W for short periods of time (PL1 limit). Sometimes this period is long enough for benchmarks to complete.

I am not sure if these TL chips have this but there is no reason why there shouldn't be any.

If my claims are true, then this may explain why TL was alot faster than Ryzen in some of the benchmarks. Technspot has an article on this power thing. Btw, anandtech benchmark the NUC10 and the 4 core i7-8559u was faster than the 6 core 10710u in some of the benchmarks.

 
Last edited:

Don Frenser

Prominent
Mar 29, 2020
27
7
535
0
Clickbait.

The headline says Intel is soooooo fast, and then the article says there is nothing really going on as things are Intel made.

How much was the author paid again by Intel?
 

deksman

Distinguished
Aug 29, 2011
218
12
18,685
0
Well, between TL having 50% higher base+boost clocks, ~20% higher IPC and most software not scaling anywhere near linearly to eight cores, it does not seem that far-fetched that a fast quad could be signfiicantly faster than a much lower-clocked octo in a fair number of cases.
TL doesn't have higher IPC at all.
Its the same as Sunny Cove... their performance enhancements come from much higher base and boost clocks because its a 4c/8th part (giving more room for higher clocks and of course more GPU cores).

Its funny how Intel conveniently forgot to mention that... or most reviewers who came out with similar titles of 'Intel could give AMD a headache'.

At any rate, Zen 3 APU's will be coming out next year, so that will actually close the single threaded performance gap (if there is one at same TDP). The only thing Intel may retain a real advantage is in iGP since enhanced Vega in Zen 3 will remain as is... however, as Zen 3 is a much stronger CPU (performance-wise) compared to Zen 2 and an entirely new uArch to boot, there's a good chance games could easily benefit from this and result in increased performance as well (so the IGP may still turn out to be the same or slightly behind XE).

Although, Intel still cannot make an 8c/16th low power CPU apparently, which makes their multi-core performance sub-par.
 
Last edited:

ASK THE COMMUNITY

TRENDING THREADS